Multi-Tenant Cloud FPGA: A Survey on Security

With the exponentially increasing demand for performance and scalability in cloud applications and systems, data center architectures evolved to integrate heterogeneous computing fabrics that leverage CPUs, GPUs, and FPGAs. FPGAs differ from traditional processing platforms such as CPUs and GPUs in that they are reconfigurable at run-time, providing increased and customized performance, flexibility, and acceleration. FPGAs can perform large-scale search optimization, acceleration, and signal processing tasks compared with power, latency, and processing speed. Many public cloud provider giants, including Amazon, Huawei, Microsoft, Alibaba, etc., have already started integrating FPGA-based cloud acceleration services. While FPGAs in cloud applications enable customized acceleration with low power consumption, it also incurs new security challenges that still need to be reviewed. Allowing cloud users to reconfigure the hardware design after deployment could open the backdoors for malicious attackers, potentially putting the cloud platform at risk. Considering security risks, public cloud providers still don't offer multi-tenant FPGA services. This paper analyzes the security concerns of multi-tenant cloud FPGAs, gives a thorough description of the security problems associated with them, and discusses upcoming future challenges in this field of study.